Google Analytics 4

BigQuery export is available for free on standard GA4 properties -- a genuine differentiator from Universal Analytics, where BigQuery export required the paid GA360 tier -- letting you run raw event-level SQL analysis beyond what the GA4 interface reports natively.

Practical Details

Linking Google Ads enables conversion import and audience sharing between the two products, and linking Search Console brings organic search query data into GA4 reporting -- both are explicit linking steps in property settings, not automatic just because you use both products under the same Google account.

Actionable Considerations

1. Enable BigQuery export if you need raw event-level analysis beyond GA4's standard reports, since it's free and often underused
2. Link Google Ads explicitly for conversion import and audience sharing, rather than assuming shared login handles this
3. Link Search Console to bring organic query data into your GA4 reporting rather than checking two separate tools

Predictive metrics (purchase probability, churn probability, predicted revenue) are GA4's concrete, named AI-driven capability -- these require a minimum volume of both positive and negative examples in your data before they activate, so a low-traffic property won't have access to them regardless of configuration.

Practical Details

The Insights panel separately surfaces automated anomaly detection (unusual spikes or drops in a metric), which is a distinct feature from predictive metrics and doesn't have the same data-volume gating.

Actionable Considerations

• Check whether your property has enough historical positive and negative examples before expecting predictive metrics to populate
• Use the Insights panel for anomaly detection specifically, since it works independently of the predictive-metrics data threshold
• Don't build a predictive-audience strategy around metrics that haven't yet activated for your property

Implementation centers on Data Streams (one per platform) and the Google tag, with Enhanced Measurement handling common interactions automatically -- custom events and conversions beyond what Enhanced Measurement covers need to be configured explicitly, either via gtag.js code or Google Tag Manager triggers.

Practical Details

One Google tag can connect to multiple GA4 properties, which is useful for sending the same event data to a main property and a test property simultaneously, but also a common source of duplicate-data confusion if not configured deliberately.

Actionable Considerations

1. Mark custom events as conversions explicitly in GA4's interface, since Enhanced Measurement doesn't automatically flag your business-specific conversion events
2. Use Google Tag Manager for implementing custom events if your team needs to iterate without code deploys
3. Audit which properties a given Google tag sends data to, to avoid unintentional duplicate reporting

Standard (free) GA4 properties have data thresholds and sampling behavior that kick in at high event volumes and high-cardinality reports -- this is the real scaling constraint, not a vague "grows with you" claim; GA4 360 (the paid enterprise tier) raises these limits and adds higher API quotas.

Practical Details

Thresholding can suppress data in specific reports (particularly ones combining many dimensions) even on properties with high overall traffic, since it's tied to data volume in that specific query, not just total site traffic -- a surprising gotcha for teams assuming "high traffic means no data limits."

Actionable Considerations

• Watch for thresholding warnings in high-cardinality custom reports, since standard GA4 applies limits regardless of overall site traffic
• Use BigQuery export for full, unsampled raw event data if standard report thresholding is a real limitation for your analysis
• Evaluate GA4 360 specifically if you need higher API quotas or need to eliminate thresholding at genuinely high volumes

"Engaged sessions" and "engagement rate" replaced Universal Analytics' bounce rate with a genuinely different definition -- a session counts as engaged if it lasts 10+ seconds, includes a conversion event, or includes 2+ pageviews, which is a specific, concrete redefinition, not just a renamed version of bounce rate.

Practical Details

Because the metric definitions changed, comparing GA4 engagement numbers directly against historical UA bounce-rate numbers is comparing two different things -- teams carrying over UA benchmarks into GA4 reporting without adjusting expectations often misread the data.

Actionable Considerations

• Don't directly compare GA4 engagement rate against historical UA bounce rate -- they measure different things
• Understand the specific 10-second/conversion/2-pageview criteria when interpreting engaged sessions
• Set new baselines in GA4 rather than assuming UA-era benchmarks still apply

GA4 replaced Universal Analytics (which was sunset in July 2023) with a fundamentally different event-based measurement model, rather than UA's session/hit-based model -- every interaction is an event, not a pageview-plus-events hierarchy, which changes how you think about reporting from the start.

Practical Details

Setup means creating a Data Stream for each platform (web, iOS, Android) and installing the Google tag (gtag.js) or configuring it through Google Tag Manager -- Enhanced Measurement, which automatically tracks scrolls, outbound clicks, site search, and video engagement, is on by default per stream and worth reviewing rather than assuming it matches your exact needs.

Actionable Considerations

1. Set up a Data Stream per platform (web, iOS, Android) rather than expecting one stream to cover everything
2. Review Enhanced Measurement's default settings, since it's on automatically and may track more (or less) than you actually want
3. Use Google Tag Manager instead of hardcoded gtag.js if you'll need to manage tags without code deploys later

GA4 uses account- and property-level roles -- Administrator, Editor, Marketer, Analyst, and Viewer -- which is a distinct role set from Universal Analytics' older permission model, so admins migrating mental models from UA should confirm the new role names map to what they actually intend to grant.

Practical Details

Data-sharing settings (like sharing data with Google for product improvement or with Google support for troubleshooting) are separate toggles from user roles, controlled at the account level, and worth reviewing independently of who has Editor or Viewer access.

Actionable Considerations

1. Review GA4's specific role names (Administrator, Editor, Marketer, Analyst, Viewer) rather than assuming they map directly to UA's older roles
2. Check account-level data-sharing settings separately from user permissions, since they're configured independently
3. Assign the Analyst or Viewer role by default for most stakeholders, reserving Editor/Administrator for those who actually configure the property

DebugView is the real-time diagnostic tool for confirming events are firing correctly as you test, distinct from the standard reports, which have a genuine processing delay (typically up to 24-48 hours for finalized data, though some data appears faster in real-time and intraday views).

Practical Details

Duplicate events are a common, specific issue caused by having both gtag.js and Google Tag Manager firing the same tag, or a tag firing multiple times on a single-page-app route change -- DebugView is the tool to catch this before it distorts your reports.

Actionable Considerations

1. Use DebugView to verify event firing during setup and testing, before relying on standard (delayed) reports
2. Check for duplicate tagging (gtag.js plus GTM both firing) if event counts look inflated
3. Expect a processing delay before data appears in finalized standard reports, rather than assuming a gap means tracking is broken